With any groundbreaking new technology, the pace of adoption climbs quickly. Over the past two decades, new platforms and tools, from the iPhone to TikTok, have seen progressively faster adoption rates. The adoption rate of ChatGPT, the artificial intelligence (AI) large language model owned by OpenAI, is unlike anything we have ever seen. Within five days of release, the platform had a million unique users.
The spike in new users has led to a deluge of thought pieces and discussions about the future of work in an AI-powered world. The world has been turned upside down with claims that ChatGPT signals the proper start of the AI age. The speedy embrace of AI tools is evident in emerging markets, where companies and governments have almost tripped over themselves to underscore their use of AI. The glaring problem is that platforms like ChatGPT are still in their infancy. While it’s clear many are eager for the AI future to take hold, the fact is that the technology isn’t anywhere close to realizing accurate intelligence or reason.
The human mind has an uncanny ability to use a small amount of data to create thoughts, language, and the ability to reason. Think about the development of language in a baby. Babies develop language with a few cues from family and the surrounding environment. This is a simplified take on language development, but it’s vital for understanding the limitations of ChatGPT when it comes to mimicking human thought. Unlike humans, large language models such as ChatGPT analyze massive data sets and produce content based on guesses about trends in the data it can access. The human mind could never process the data required to make ChatGPT function, nor would it require so much data.
At the heart of OpenAI’s approach to ChatGPT is the notion that human behavior is predictable. By analyzing large data sets, OpenAI’s algorithm can essentially guess what we are thinking or looking for in an answer. While we might have a knee-jerk reaction to the claim that human behavior is predictable, we must consider the effect of nearly two decades of internet and smartphone usage. Our behavior is increasingly predictable because we have allowed algorithms to shape what we read and digest on the internet. Features like Google’s autocomplete in Gmail can accurately predict what we want to write with a startling degree of accuracy. Moreover, the direction of travel is evident as we spend more and more time on our phones. When was the last time you visited a bookstore or a library and serendipitously stumbled upon a book you didn’t know you were looking for?
While we might be becoming more predictable, AI tools still lack the fundamental ability to reason, as the linguist Noam Chomsky recently explained in the New York Times: “Whereas humans are limited in the kinds of explanations we can rationally conjecture, machine learning systems can learn both that the earth is flat and that the earth is round. They trade merely in probabilities that change over time. For this reason, the predictions of machine learning systems will always be superficial and dubious.”
This shouldn’t undermine the incredible value of ChatGPT and other AI tools. They use impressive computing power to harness our increasing online predictability. Yet the overzealous embrace of these technologies completely misses this vital point about human development and its impact on society.
There are also many other ethical issues in the quick embrace of these technologies. In its latest white paper for the launch of ChatGPT 4, OpenAI is quick to note that “AI systems will have even greater potential to reinforce entire ideologies, world views, truths, and untruths, and to cement them or lock them in, foreclosing future contestation, reflection, and improvement.” This is scary stuff, especially in light of recent developments like Microsoft’s dismissal of its entire AI ethics team and the general lack of ethics parameters in AI development.
The ethics debates will only grow in light of a critical point raised by Chomsky that ChatGPT and other AI tools are “constitutionally unable to balance creativity with constraint. They either overgenerate (producing both truths and falsehoods, endorsing ethical and unethical decisions alike) or undergenerate (exhibiting noncommitment to any decisions and indifference to consequences).”
The truly beneficial future with AI tools lies in collaboration instead of a complete takeover or outsourcing. These tools can help us produce work of all kinds more efficiently, but they can’t replace the human mind. The iPhone, for example, facilitated better communication and made the world more accessible but didn’t change the essence of how we work. ChatGPT won’t change the nature of work, either.
Companies and governments rushing to upend their working models would be well served to pump the brakes and consider more incremental AI incorporations. This will be challenging given how aggressively the technology sector pushes ChatGPT, but the fad will eventually fade and be replaced by another platform. We should use this opportunity as a society to reflect on the nature of human intelligence and how it’s being changed in the technology age.
Joseph Dana is the former senior editor of Exponential View, a weekly newsletter about technology and its impact on society. He was also the editor-in-chief of emerge85, a lab exploring change in emerging markets and its global impact. Twitter: @ibnezra